royorel
commented
3 years ago Hi @Serjura,
The functionality you're looking for is in this function: https://github.com/royorel/Lifespan_Age_Transformation_Synthesis/blob/7be4534848e94d9493dfb6da8dfd5724fa5f6766/util/preprocess_itw_im.py#L97
As for your other question. Our algorithm does not output a specific age but rather an age cluster, the naive age prediction would be to interpolate between the means of the age classes in the same ratio you interpolate between the latent code. However, we didn't do it in the paper, and we don't have any guarantees that this kind of prediction is indeed accurate.
Sergej-dcai
Hi Royorel
Really impressed by your work here! I am amazed by the accuracy of the outputs on some tests I did.
For my Python project, it would be even more amazing to be able to use the outputs combined with the original image, seeing the aging process on the photo itself basically. I tried digging around in the code, but I couldn't find a way to determine how the face is cut out of the image (basically I need to know "where" I should paste it over the original image). Is there any way you could help me out with this project?
As a side note, I was wondering if your algorithms can output the predicted age or some other way to determine the output image most fitting to the input. Is there something in the code that can give me an indication which pictures make the person older and which make the person younger?
Thanks a lot in advance for your helpful answers and keep up the astonishing work!